What would it take to implement built-in conversion to and from System.Decimal?

[siegfriedpammer]

What would it take to implement built-in conversion to and from System.Decimal?

[siegfriedpammer]

This was not a question, but rather a feature request. Also, why is the idea misguided? If the idea of automatic conversions of built-in types is so wrong, why do they exist for strings, integers, floats and all other built-in types EXCEPT decimal? Also note that the range of the type should not be a problem, because Python can use arbitrarily sized numeric types. The conversion from Python to .NET could fail with an OverflowException IMHO.

In my case I cannot even circumvent the problem by using the Equals method. x.Decimal.Equals(42) would work, because System.Decimal stays System.Decimal in Python. However, x.Int.Equals(42) does not work, because System.Int32 is marshalled to Python int and therefore has no Equals method. This causes interesting surprises when working with boxed primitive types. x.Boxed.Equals works with decimal, but fails with int.

Would be nice if guidance could be provided somehow. Thanks!

If you don't want to answer the questions here, I will be glad to open a discussion in a few days.

[lostmsu]

Generally misguided idea. You should use https://docs.python.org/3/library/decimal.html

Also not a bug. For questions please use discussions.

[lostmsu]

The particular idea is misguided because System.Decimal does not map neither to int - because Decimal has fractions, nor to float - because floats can not precisely represent decimals, and precise calculations are the whole point of decimals.

You need a compelling use case. Your last example is not very compelling either because x.Int == 42 is more idiomatic in Python, and if you want C# semantics you can use System.Object.Equals(a, b).

[siegfriedpammer]

Thanks I already implemented an encoder that maps System.Decimal to decimal.Decimal, which solved most of the issues. However, I would welcome a change that adds a value parameter to the CanEncode/CanDecode methods, so the callee has all information available.

[lostmsu]

What scenario do you need this for? CanEncode/CanDecode are only there to quickly filter out types the codec does not support. Without them Python.NET would have to go through all codecs for every single attempt to convert anything, which would kill performance.

[siegfriedpammer]

Sorry for taking so long to reply. I have missed the notification.

A value parameter would enable the following:

public bool CanEncode(Type type, object value)
{
    return type == typeof(decimal)
        || (type == typeof(object) && Type.GetTypeCode(value?.GetType()) == TypeCode.Decimal);
}

Then it would be possible to handle boxed decimals without calling TryEncode for all objects.